Thanks for your thoughtful answers to my previous questions. By using the phrase
"pat answer" I didn't mean to imply that you were giving false information; but
perhaps that the casinos, designers, and even the Gaming Commissions (well,
perhaps only in Mississippi!) were misleading us all! Your sometimes response of
"Still, only the casinos know for sure" worried me a bit; and I really did want
to believe in the RNG and having a fair chance; and that's not what I'd been
observing at the Grand and Copa Casinos in Gulfport, MS, where even the RNG's
paltry payback as statistically "modulated" by the Telnaes map sometimes seems
to be ignored.
Also the answer that Andrew Glazer gave
in his book entitled Slots! How to Improve Your Odds at Winning (2001) did not
ring true either. He claimed, e.g., that a machine is ordered by the casino
(rated and tested) to payback a certain percentage (say, 90%) but that it may
not achieve that rated payback for months or years; THUS some machines run hot
(he mentioned a machine paying back at 102% for months) and some run cold. This
had led me to the unfortunate use of the phrase "statistical aberration" and you
rightly pointed out that "local non-randomness is an essential part of
randomness." I had tested series-based RNGs for use in simulations on an IBM
7040 in 1961 at Auburn U. and would have thrown out immediately any RNG
algorithm that was as bad as Glazer (a gambling columnist for the Detroit Free
Press, at least in 2001) had indicated.
So back to square one. And you made me think about what I did know for certain:
1) I did NOT know whether the RNGs/ROMs were changed in the tournament machines
at the Grand (they probably once were, when all 20 were put out of service for a
week); 2) I DID KNOW that the players were told that the Double Diamonds were
made "hot" and that all had an equal chance at winning; and, 3) I knew that most
of the machines run quite cold most of the time since I play those machines
weekly in normal service and had detailed payback rates (usually less than 90%, in individual cases as low as 50%) and frequency (averaging about 8
plays per payback) data on them. I had caught one "hot" machine (probably not
reset after a tournament) and it paid back every 4 plays at a rate of about
So I finally found a friendly Slot Tech and I asked him how the tournament
machines were made hot AND his answer was very interesting: in his words "the
odd [physical reel] position numbers are ignored in tournament mode." Now there
are 11 blanks and 11 non-blank positions on the reels of the Double Diamond. By
ignoring blanks (or substituting a non-blank for a blank by simply bumping the
index) the odds in your favor for a Double Diamond coming up on the payline
would be doubled; and your chances of getting the triple Double jackpot would be
increased by 8. The triple Double Diamond might come up once in 1,000 plays
rather than the usual 1 in 8000 (I'm just making these numbers up since I don't
have the Telnaes maps and I'm using 20 in the calculations, not 22).
A tournament round runs for seven minutes; a payback probably occurs every four
plays (each play requiring generation of three numbers by the RNG) rather than the
average of eight or so in normal play. Thus in one round each machine would give
about 105 payoffs (I've counted many times) limited by the time to move the
reels plus the time to get to the next payoff (four PLAY 3 CREDITS presses at
about 1 second for most players). So with all machines in use the total number
of plays would be 20 x 420 = 8,400 and you might expect up to 8 triple Doubles in
the round. Sure, I know it might average out to that over the long run, and you
might get only one or two which is usually observed in a round; and these
numbers ignore how the virtual reels and Telnaes maps factor into this, and the
odds are less than 1 in 8,000. Anyone carefully observing these tournaments
would never play these machines!
One further comment: the typical total payback for a machine played optimally
during a tournament usually ranges from 2500 to 4000, which tells more about the
difference between RNGs than the players. The person who gets the jackpot (if
any) of 2500 credits usually is the winner; unless beaten by the even luckier
person who gets two triple doubles in the round . . . which occasionally happens and
more frequently than you might expect, back-to-back.
Now making the machine hot in this way is what I would call "logic above the
RNG" and opens up many other possibilities which I had been refusing to believe;
but, as a programmer, could easily envision. You answered one of those: it's
illegal to change the Telnaes maps under computer control (at least in most
jurisdictions). Well, the maps weren't changed to make the tournament machines
hot. Could it be that a simple "control parameter" like this could be set by the
network; defaulting to the "cold" mode if the network or cluster controller were
down? I assume the Gaming Commissions have many expert programmers on their
staffs who would carefully analyze the control programs for such "logic above
the RNG" tricks? Is that a good assumption?
One other nagging question lingers; a programming technique that would also
qualify as "logic above the RNG." MacIntyre Symms in his book Slot Machine
Strategy: Winning Methods for Hitting the Jackpot (2004)) defines "pay cycle" as
a period when a machine is under its rated payback percentage and must pay back;
and "take cycle" as the opposite. Administering this would require overriding
the RNG at times. Is this actually done? And legal?
Thanks again for assisting me in developing my notes for a course in Slots
First, let me apologize for not responding sooner. I try to keep these "conversations" going by answering in a timely fashion, but your reply slipped through the cracks. Your original letter appeared about six months ago.
Now, let me address some of the statements in your letter.
You mention the "RNG's paltry payback." I just want to emphasize that RNGs do not have paybacks. All the RNG does is generate a series of numbers. It's the layout of the symbols on the virtual reel that determines the payback. I emphasize this point only because people sometimes think some RNGs are more generous than others, and the RNG really has nothing to do with determining the payback.
As for throwing out the RNG algorithm used in the machine Glazer described, you're confusing the performance of the machine (how quickly a machine approaches its long-term payback) with the performance of the RNG. You have to look at the numbers generated by the RNG to determine how well the RNG performs, not the payback of the machine. I'm skeptical that a machine paid back 102% for months, but Andy wouldn't have written it if it hadn't happened. We have to remember that just because an event is unlikely, that does not it is impossible.
I have had mixed experiences getting good answers from slot techs. I think a key point to remember is that techs are trained in the hardware of the slot machine and how to use the software running it. In most cases, they are not programmers. They may not necessarily know how the programmers have achieved certain results. I suspect what the programmers really did is create a set of virtual reels that do not have any blanks on them. This is a piece of cake, unlike altering the program to ignore the odd physical reel positions.
And I think this response indicates that the tech is not a programmer. How is the program to ignore those positions? What does it do if the number from the RNG indicates a blank should land on the payline? Get another result from the RNG or increment the physical stop to an even-numbered one? Ignoring the odd stops may be the effect, but he didn't say how that was achieved.
Any "logic above the RNG" is illegal in the United States. The more common term for this sort of thing is "secondary decision." The result indicated by the output of the RNG is what must be displayed — period.
Those definitions of "pay cycle" and "take cycle" are pure BS. Machines run hot and cold. That's true. But it is purely the result of random selection of outcomes. Overriding the RNG is illegal.
Best of luck in and out of the casinos,
Just a few quick questions that came up as I read your recent articles:
- Are there any slots that use hardware RNG chips, or do they all use pseudo-RNGs that are started with a seed number and are deterministic?
- On a pseudo-RNG machine when will the sequence start over (when powered down, or reset by an attendant)?
- You mentioned the tournament machines — I had e-mailed you once before about the Double Diamond tournament machines at the Grand Casino in Gulfport, MS (no longer there and won't return! Although 3 casinos have reopened in Biloxi after Katrina) . . . I had speculated that it was perfectly legal (at least in Mississippi) to play the tournament machines, and that they were made "hot" by throwing a switch, and that an attendant had told me that in the "hot" mode the simple expedient of ignoring blanks caused them to be hot.
Thanks for your assistance. Hurricane Katrina has pretty well cured me of my gambling addiction!
I finally answered the e-mail you referred to in #3 above.
I don't know of any manufacturer that uses a hardware-based true RNG, but that doesn't mean that no one does. The manufacturers don't reveal many details about their RNGs. And I won't press them for an answer on this issue because it really doesn't matter.
The number of numbers an RNG can generate before it repeats is known as its period. Kilby and Fox, in Casino Operations Management, wrote that IGT's RNG generates over 4 billion numbers before repeating. That assumes that the RNG is not reseeded. A modern machine will reseed the RNG at certain times — exactly when is another detail manufacturers are reluctant to reveal.
Send your slot and video poker questions to John Robison, Slot Expert, at email@example.com. Because of the volume of mail I receive, I regret that I can't send a reply to every question. Also be advised that it may take several months for your question to appear in my column.